Punctuation and Typography in Translation

How to translate punctuation and typography without changing meaning

How should translators handle quotation marks, commas, dashes, ellipses, brackets, lists, line breaks and right-to-left text without accidentally changing meaning? The answer is to distinguish semantic punctuation from language-specific typographic conventions, identify each mark’s scope and function, then verify the result in the actual reading environment. A comma that sets off a parenthetical is not merely a decorative mark, and a dash that introduces a condition cannot safely disappear just because another language prefers different typography. This comprehensive guide explains how to translate punctuation accurately, keep quotation boundaries intact and preserve the logic of written text across languages.

The hidden problem is that text is simultaneously language, structure and display. A translated sentence can use excellent vocabulary while presenting a quotation as the author’s own assertion, changing which list item an exception applies to, or making an address or reference code appear in the wrong order. A line break in a poem may carry rhythm or opposition; a line break in a safety instruction may determine which action is associated with which warning. In right-to-left interfaces, the visual sequence of mixed-script text can differ from its logical stored order. Neither fluency nor spell-checking alone proves that a translated page is faithful.

Readers searching for translation punctuation rules, how to preserve quotation marks in translation, how to localise lists and ellipses, how to handle multilingual line breaks or why Arabic and English text sometimes display punctuation incorrectly need a repeatable method. This chapter separates meaning-bearing boundaries from locale conventions, documents the role of Unicode text algorithms and W3C markup guidance, walks through common failure modes and provides fictional worked diagnostic cases. The objective is practical: a reader should understand who said what, what is omitted, which condition belongs to which instruction and what sequence of text to read regardless of language or device.

A quick working rule: first recover the logical text, then identify semantic boundaries, then identify punctuation conventions, then produce idiomatic target-language typography, and finally test rendering, selection, copying, accessibility and export. Do not confuse a character’s Unicode identity with how it appears on screen. A visible glyph, a stored code point, a semantic separator and a layout break may be related but are not interchangeable.

This is a specialist architecture chapter, not a replacement for the existing introductory page on how to translate punctuation, quotation marks and capitalisation naturally. That article remains the entry point for ordinary sentence editing. The present chapter handles the broader cross-document integrity problem: quotation ownership, annotation scope, list-item dependencies, paragraph and line semantics, bidirectional display, grapheme clusters, Unicode line breaking, structured extraction and multi-format testing. It is also not an alternative to dedicated W3C language and direction metadata guidance, which owns its separate technical topic.

This article uses invented examples as editorial diagnostics, not claims about real people or incidents. There is no universal rule requiring every language to use English punctuation. Typographic practices vary by script, language, style guide, publishing medium and genre. For legal, medical, financial or safety-critical instructions, a qualified domain reviewer should check whether punctuation changes alter duties, conditions, identity, quantities or scope. The goal is to preserve the communicated relation, not to impose one punctuation inventory worldwide.

The problem hidden in the page

Consider “Do not release the equipment; inspection pending.” A translator might turn this into two headings, a single imperative or a list with one stray subordinate line. The essential distinction is between a prohibition and the reason it remains active. A reader must not interpret the second clause as a completed inspection or the first as a conditional suggestion. The semicolon is only one means of signalling that relationship; the target can use a full stop, conjunction, typographic panel or another conventional form if the action boundary remains intact.

Now consider a reviewer’s words “The material is safe,” followed by a sentence saying the quote is disputed. If the quotation marks are dropped during translation or the quote is folded into a headline, a third-party claim becomes the document’s own conclusion. This may seem like a simple copyediting mistake, yet it is a shift in evidentiary ownership. Good typographic translation therefore begins by asking which marks govern quotation, which govern sentence structure and which only govern conventional appearance.

A third example is a numbered list whose final item reads “except where prior approval is required.” If a renderer pushes that phrase onto the next line without its bullet marker, readers can attach the exception to the entire list. That is not a spelling error. It changes logical scope. The cure is to model the exception explicitly in source structure, associate it with the intended item and verify the visual hierarchy after localisation.

The same issue emerges in digital documents with invisible controls or direction changes. A number in an Arabic- or Hebrew-script sentence may be displayed left to right inside a right-to-left passage. Punctuation adjoining that number can be visually displaced unless the run is isolated appropriately. The stored character sequence and the rendered image serve different functions. A translator must know which layer the error occupies before trying to repair it. Manually reversing source characters to make one screenshot look right can create worse problems in copying and accessibility.

What Unicode and W3C do—and do not—guarantee

Unicode Standard Annex #9, the Unicode Bidirectional Algorithm, specifies how text containing right-to-left and left-to-right characters is ordered for display. The Unicode 18.0.0 revision was published on 1 September 2026. Its algorithm is not a translation engine, and it cannot know whether a quoted sentence has been translated faithfully. It supplies a rendering model that interoperable software can apply to a logical character stream.

Unicode Standard Annex #14, the Line Breaking Algorithm, also updated for Unicode 18.0.0 on 1 September 2026, identifies default line-break opportunities and mandatory behaviours. It does not determine exactly where every application should wrap a paragraph. Width, font metrics, layout constraints and tailoring can influence actual line endings. A line-break opportunity is therefore not always a semantic sentence boundary, just as an automatically wrapped line is not a new paragraph.

Unicode Standard Annex #29 provides default grapheme, word and sentence segmentation. Its September 2026 revision helps implementers understand user-perceived character clusters and boundaries. A displayed letter plus combining marks may occupy several code points while functioning as one grapheme cluster. Translators reviewing truncation, copy-and-paste, cursor movement or character-count limits need to distinguish bytes, code points and graphemes. These units answer different questions, and no one counting system fits all scripts.

The W3C Internationalization guidance on inline bidirectional markup recommends using appropriate direction markup, including isolation, for embedded runs whose base direction differs from surrounding text. In ordinary HTML, semantically placed direction attributes and the bdi element are preferable to reversing characters or inserting untracked invisible controls merely to make a single display look correct. Where markup cannot be used, Unicode isolate characters may be relevant; their introduction requires engineering review and regression tests.

W3C’s international typography material shows that languages and scripts use different approaches to line breaking. Some break primarily between space-delimited words; others permit different character- or syllable-related breaks and impose constraints around punctuation. A style guide is needed for language-specific typographic decisions. The Unicode algorithms are infrastructural defaults, not full descriptions of the literary or legal punctuation practices of every community.

A typography meaning ledger

Ledger field 1: Claim owner

Who actually asserts, asks, doubts or quotes each proposition? Quotation punctuation may separate source speech from commentary. If lost, accountability shifts.

In production, record the source span, its logical function, the intended target-language representation and the test environment in which the output will be checked. This small amount of explicit documentation prevents language reviewers from judging typography only by how familiar it looks. When a layout change cannot preserve the intended boundary, raise a source-author query rather than silently reassigning scope.

Ledger field 2: Scope boundary

Which word, phrase, clause or list item is modified by the warning, exception, negation or condition? Bullets and indentation may carry logical relations that punctuation alone cannot express.

Ledger field 3: Omission marker

Does an ellipsis indicate omitted words, interrupted speech, trailing thought or layout truncation? Different editorial purposes require different explanatory conventions.

Ledger field 4: Hierarchy

Which text is a heading, a step, a substep, a note, a footnote or a table header? Visual indentation cannot substitute for stored structural hierarchy.

Ledger field 5: Display direction

What base direction and embedded text runs apply to each segment? Mixed-script output may rearrange punctuation and numbers if boundaries are not isolated.

Ledger field 6: Temporal and numeric tokens

Are separators part of dates, decimal values, ranges, version codes or identifiers? Replacing a separator without classifying the token can change a quantity or identity.

Ledger field 7: Reading route

What sequence should a person or assistive technology follow? A visually plausible page can have a wrong extraction or reading order.

Ledger field 8: Genre and medium

Is this text dialogue, poetry, a contract, a help panel, a database field or a subtitle? The same graphic mark has different communicative duties in different genres.

Ledger field 9: Local convention

What punctuation, quotation style, spacing or typography is appropriate for the target language and editorial house? Conventional differences can be implemented without modifying semantic boundaries.

Ledger field 10: Version relationship

Did the source punctuation itself change between revisions? The target may need a new review even when its dictionary words did not change.

Thirty-eight punctuation and layout mechanisms

1. Direct quotation

Meaning mechanism: Opening and closing quotation delimit speech or cited text belonging to an identified originator. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Use target-language quotation conventions while preserving the source boundary and attribution. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A claim outside the quotation may otherwise become part of the quoted speaker’s words. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

A practical repair has three parts. First identify the smallest span whose boundary carries the meaning. Second reconstruct the same boundary in the target through punctuation, syntax or semantic markup as appropriate. Third verify the result in the rendered environment and in copy-and-paste or screen-reader order where relevant. Keep this test separate from spelling and dictionary review; correctness at one layer does not guarantee correctness at another.

2. Nested quotation

Meaning mechanism: Several speakers or sources can be quoted within one passage. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Maintain quotation nesting with language-appropriate delimiters or explicit attribution. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Flattening levels can put an interviewer’s paraphrase in a witness’s mouth. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

3. Scare quotes

Meaning mechanism: Quotation signs can distance a writer from an expression without quoting actual speech. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Distinguish ironic or contested terminology from verified direct quotation. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Adding or removing irony changes the writer’s attitude. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

4. Cited titles

Meaning mechanism: Book article product and artwork titles may be italicised quoted or set apart under local convention. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Protect the title as an entity without implying it is dialogue. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A title can otherwise be misread as a claim in the surrounding sentence. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

5. Apostrophe and elision

Meaning mechanism: An apostrophe may mark possession contraction elision or transliteration conventions. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Check the morphological relation rather than copying every visible apostrophe. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Wrong replacement can change a personal name or possessive relationship. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

6. Comma in restrictive clause

Meaning mechanism: Punctuation can signal whether a phrase identifies the intended member of a group. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve whether the modifier selects a subset or adds a side comment. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A non-restrictive reading may broaden the set of people covered. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

7. Comma in subordinate clause

Meaning mechanism: Commas can indicate syntactic attachment and avoid garden-path readings. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Write an unambiguous target clause with native punctuation. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Moving a comma may attach a warning to the wrong action. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

8. Semicolon joining clauses

Meaning mechanism: A semicolon may signal related independent statements without a full connective. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep the relationship explicit if target punctuation differs. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: An accidental list break can detach a reason from its prohibition. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

9. Colon as announcement

Meaning mechanism: A colon can introduce explanations definitions citations or enumerations. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Determine what material is introduced and whether it is exhaustive. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Changing colon scope can turn examples into requirements. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

10. Dashes as parenthesis

Meaning mechanism: A dash pair may isolate an aside or qualification. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep the aside separate in words or punctuation. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A single dash may turn a qualification into the main conclusion. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

11. Dash as interruption

Meaning mechanism: A dash may represent an interrupted or abruptly abandoned utterance. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve interruption rather than converting it to completed speech. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Reader may treat an interrupted promise as fully made. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

12. Hyphen as word formation

Meaning mechanism: Hyphenation may signal a compound modifier or conventional written word. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Consult morphology and style for the target language. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A misplaced hyphen can reverse what a descriptor modifies. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

13. Minus sign versus hyphen

Meaning mechanism: The mathematical minus sign and an orthographic hyphen are different characters and roles. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve mathematical operators and correct spacing independently of local prose style. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A negative value may become a name or label. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

14. En dash as range

Meaning mechanism: A range marker relates endpoints and often compresses inclusive or approximate spans. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Verify mathematical meaning before adopting local range typography. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Two endpoint numbers can become subtraction or a list. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

15. Em dash as turn marker

Meaning mechanism: Some dialogue conventions use dashes to introduce speakers. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve speaker turns even if target convention uses quotes. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Dialogue may merge voices when marks are normalised. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

16. Ellipsis indicating omission

Meaning mechanism: An editorial ellipsis may stand for material removed from a quotation. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep the omission transparent and avoid inventing continuity. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Readers may attribute an edited statement verbatim. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

17. Ellipsis indicating hesitation

Meaning mechanism: Three dots can represent a pause or trailing thought. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve discourse function instead of treating every ellipsis as deletion. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A doubtful answer may become a firm response. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

18. Ellipsis in interface

Meaning mechanism: Truncation or a menu label with an ellipsis may signal further interaction. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Distinguish a visible truncated string from a control that opens a dialogue. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: The user might think an operation has already executed. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

19. Parentheses as aside

Meaning mechanism: Parentheses often mark supplementary context rather than core operative content. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve what is optional background and what is binding instruction. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Moving content out of parentheses may promote it to a requirement. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

20. Parentheses as alternatives

Meaning mechanism: Brackets can encode optional suffixes or alternative constructions. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Disambiguate which option a reader is authorised to choose. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Readers may complete mutually exclusive choices together. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

21. Square brackets editorial

Meaning mechanism: Inserted words in a quotation can signal editorial clarification. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep inserted text distinct from original speech. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A translator’s addition can become an alleged verbatim claim. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

22. Brackets in references

Meaning mechanism: Citation brackets may identify sources footnotes or numbered materials. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve reference identity and ensure targets resolve. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A source link can detach from the claim it supports. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

23. Full stop and sentence boundary

Meaning mechanism: Punctuation may separate obligations definitions and explanatory remarks. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Map proposition and scope before deciding sentence count. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Merging sentences can change whether a condition governs both. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

24. Question mark and speech act

Meaning mechanism: Question punctuation often indicates a request for information rather than a statement. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve actual interrogative force and implied alternatives. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A rhetorical question can be misread as a direct assertion. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

25. Exclamation and intensity

Meaning mechanism: A mark may express urgency emotion or conventional salutation. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Match pragmatic force without fabricating an emergency. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Exaggerating punctuation can inflate risk. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

26. Enumerated lists

Meaning mechanism: Numbered items may encode sequence priority or legal structure. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep item identity and any cross-references stable. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Reordered numbers can change the requested action. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

27. Unordered lists

Meaning mechanism: Bullets can indicate independent options or cumulative requirements. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Decide whether and or or applies and clarify where needed. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A shopping list can be mistaken for a procedural checklist. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

28. Nested list hierarchy

Meaning mechanism: Indentation and sublists determine which item governs a subcondition. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Use native structural lists not visual spaces alone. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: The exception can attach to an entire parent list. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

29. Table cell boundary

Meaning mechanism: Rows and columns provide relationships not conveyed by prose punctuation. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Keep headers units and footnotes attached to values. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A shifted cell can report an incorrect owner or amount. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

30. Heading punctuation

Meaning mechanism: A heading ending with a colon may introduce only the next section. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Retain section boundaries in semantic markup. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A subsequent condition may be read as globally applicable. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

31. Paragraph break

Meaning mechanism: Paragraph structure can separate a rule from a later explanation. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Preserve logical grouping even when layout changes. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: The reason for a rule may be misread as an additional rule. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

32. Hard line break

Meaning mechanism: A deliberately inserted break can control poetry addresses or scripted dialogue. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Mark semantic line breaks separately from automatic wrapping. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A mailing address or verse may change sequence. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

33. Soft wrap

Meaning mechanism: Automatic wrapping is presentation not new text structure. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Do not insert source newline tokens for every rendered line. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Responsive layout becomes brittle and extraction gains false breaks. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

34. Whitespace grouping

Meaning mechanism: Spaces may distinguish tokens in codes or form input from ordinary words. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Identify whether whitespace is semantic or cosmetic. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: Normalisation can corrupt a reference code. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

35. Nonbreaking space

Meaning mechanism: Some units and constructions must not be divided across lines. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Use appropriate no-break behaviour while keeping text accessible. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A unit or figure may separate from its quantity. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

36. Grapheme cluster

Meaning mechanism: A perceived written character may contain multiple Unicode code points. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Protect grapheme boundaries during truncation and selection. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A combining mark may detach and distort a letter. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

37. Bidirectional isolate

Meaning mechanism: Embedded opposite-direction text must not reorder surrounding punctuation. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Use markup-based isolation where the host format permits. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: An identifier or closing parenthesis may appear displaced. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

38. Font-dependent glyph

Meaning mechanism: The same code point can be rendered differently by fonts and styles. The translator’s responsibility is to identify whether this relationship is expressed by a symbol, syntax, formatting convention or document structure in the original. Merely reproducing the same visible punctuation does not always reproduce the same communicative effect. Conversely, automatically replacing every symbol according to a house-style preference can destroy a relationship that the source deliberately marks.

Target-language decision: Inspect both character identity and appearance. Read the surrounding sentence, not the punctuation mark in isolation. Where the target language needs a different delimiter, rearranged clause or additional word to make the same distinction, select the natural construction and record why it is equivalent. Where an identifier or quoted original must remain unchanged, separate untranslated data from translated framing instead of rewriting the data as prose.

Likely failure: A font substitution may mislead readers or hide marks. To detect it, remove the source page from the reviewer’s sight and ask what the translated page alone permits them to conclude. Compare that interpretation with the source text’s actual scope, ownership and boundaries. If the answer changes who said something, which action is required, or what data string is represented, the typography has become a meaning error.

A twenty-stage punctuation integrity workflow

Gate 1: Set the translation brief

List languages scripts user needs output formats and the documents in which punctuation changes could affect action. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Run a contrastive test before and after the proposed edit. Ask whether changing the mark could reasonably change the underlying proposition, speaker ownership, selection of an item, or the behaviour of a software renderer. If yes, route the change through language and technical review; if no, apply the target’s normal editorial convention and move on. Preserve the rationale so later localisation updates do not recreate the defect.

Gate 2: Read the complete source

Identify source-author intent around quotations lists headings tables examples and notes before editing tokens. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 3: Extract logical text

Separate stored order from a screenshot or PDF visual arrangement and preserve original character identity where possible. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 4: Mark semantic spans

Annotate quote boundaries speaker turns restrictions conditions and subordinate clauses as structural relations. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 5: Separate style from meaning

Classify each mark as semantic grammatical bibliographic conventional numeric or purely decorative. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 6: Map source quotation owners

Identify whose speech begins and ends and how nested or partial quotations are signalled. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 7: Map list and table scope

Connect every exception subordinate item unit note and definition to its intended parent. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 8: Identify data tokens

Protect codes paths negative signs decimal separators time formats mathematical operators and stable identifiers. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 9: Set target conventions

Use verified punctuation and typography conventions for the target language script and genre. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 10: Draft language first

Express the intended meanings naturally without forcing source language punctuation one for one. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 11: Reinsert structural markup

Render headings lists quotations footnotes and tables as native structure rather than typographic imitation. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 12: Handle mixed direction

Set direction metadata at proper block and inline boundaries and isolate embedded runs when appropriate. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 13: Review Unicode segmentation

Check grapheme-safe editing and line-break opportunities against target script requirements. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 14: Inspect display metrics

Test readable fonts text expansion wrapping collision and punctuation around embedded strings. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 15: Validate logical extraction

Copy text and compare source order with the order provided to search accessibility and export systems. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 16: Back-read a source-free version

Ask a reviewer to identify who said what what was omitted and what conditions apply using only target output. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 17: Check responsive variants

Review a narrow phone width wide desktop print PDF and accessibility reading routes as appropriate. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 18: Run regression cases

Retest marked quotes lists negative numbers mixed scripts and codes after every significant source or UI change. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 19: Document unresolved ambiguity

Query the source author when a punctuation boundary permits materially different meanings. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Gate 20: Approve and archive

Store source revision target revision rationale reviewer and reproducing test cases before release. This gate turns an aesthetic editing preference into a testable translation decision. Annotate the relevant source example and specify what a target reader must still be able to reconstruct. A check that simply marks the result “looks fine” is insufficient whenever punctuation influences obligation, attribution, quotation, reference identity, numerical interpretation or navigational order.

Thirty-five failure modes and their practical repairs

A failed punctuation translation can arise at several layers: the translator may misunderstand the sentence, the publishing system may remove markup, a font may disguise a character, a content editor may merge list levels, or a right-to-left renderer may rearrange neutral punctuation. Diagnose the layer before fixing it. The examples below concern the meaning a reader receives and include the smallest corrective action that can restore the intended boundary.

Failure 1: Quotation boundary lost

Symptom: The quoted witness statement becomes the journalist’s own claim. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Restore the quote frame or explicit speech attribution. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Who says the proposition? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 2: Quotation boundary expanded

Symptom: An editor’s warning is accidentally placed inside a quote. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Move the warning outside and preserve the correct speaker. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which words are verbatim? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 3: Embedded quote level flattened

Symptom: A speaker quoting a colleague seems to speak the colleague’s words personally. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Retain both speaking layers. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Who originated each statement? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 4: Editorial insertion unmarked

Symptom: Added words inside a quotation appear authentic. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Use a transparent editorial convention or a paraphrase. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Can the reader distinguish source from clarification? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 5: Scare quotes normalised away

Symptom: A contested label appears endorsed. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve distance where rhetorically intended. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does the speaker accept that label? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 6: Comma changes modifier scope

Symptom: A restrictive condition becomes a nonrestrictive aside. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Rebuild the intended subset in target syntax. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which people or objects are included? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 7: Semicolon treated as list separator

Symptom: A second complete statement becomes an independent option. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Keep clause relation explicit. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Is the reader offered a choice? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 8: Colon changes requirement count

Symptom: Examples following a colon are mistaken for exhaustive conditions. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Clarify whether the set is illustrative or complete. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does the list define all possibilities? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 9: Dash changes speaker turn

Symptom: A dialogue dash becomes a mathematical minus or parenthesis. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Restore the conversational speaker boundary. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Who is speaking now? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 10: Interrupted sentence completed

Symptom: A trailing dash is edited into a full promise. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Keep the abandoned or interrupted utterance visible. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: What was actually committed? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 11: Ellipsis of omission removed

Symptom: A shortened quotation appears complete. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Indicate editorial deletion under the style guide. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Was text left out of this quotation? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 12: Hesitation treated as deletion

Symptom: A conversational pause appears as an editorial cut. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve the pause with appropriate discourse wording. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Did the speaker omit words or hesitate? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 13: UI ellipsis lost

Symptom: A menu choice that opens a dialogue appears to act immediately. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve interaction expectation where the interface convention requires it. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: What does activation do? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 14: Two list levels merged

Symptom: A narrow substep is elevated into a general rule. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Restore semantic list hierarchy. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which parent item governs the substep? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 15: List exception detached

Symptom: An exception appears to apply to all options rather than one. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Place the exception under the correct item. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which item has the exception? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 16: Sequence renumbered

Symptom: A step referenced by number loses its identity. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Update numbering and cross-references together. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does the same step remain referenced? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 17: Bullet treated as conjunction

Symptom: Independent choices become cumulative requirements. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: State whether one or all items apply. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Is the reader expected to do every item? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 18: Table header shifted

Symptom: A value is read against a different column. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve structural table semantics and inspect the render. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which column defines this value? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 19: Footnote anchor misplaced

Symptom: A qualification attaches to the wrong claim. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Reattach source and note anchors. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which statement does the note limit? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 20: Text direction guessed wrongly

Symptom: Neutral punctuation moves around an embedded RTL title. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Set appropriate direction and isolate the embedded run. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: What is the logical character order? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 21: Manual character reversal

Symptom: Source order is changed to match a screenshot. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Restore logical stored order and use rendering controls. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does copying the text return the right order? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 22: Invisible controls untracked

Symptom: Direction controls survive into a different host system unexpectedly. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Prefer visible markup or document each control. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which invisible characters exist? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 23: Code point split

Symptom: A grapheme is truncated between a base and combining mark. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Use grapheme-aware editing and truncation. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Is the displayed character intact? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 24: Nonbreaking sequence broken

Symptom: Quantity and unit separate across distant lines. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Apply safe wrapping constraints or rewrite grouping. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does the value remain paired with its unit? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 25: Soft break exported hard

Symptom: Automatic wrapping becomes sentence or field separation. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Keep layout wraps separate from content breaks. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Did export add new textual structure? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 26: Hard break deleted

Symptom: Poetic or address lines merge. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve intentional line units with semantic markup. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Are the original line relationships visible? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 27: Negative sign substituted

Symptom: A minus becomes a hyphen or vanishes. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve numerical sign and parse value. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Is the quantity negative? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 28: Range delimiter misunderstood

Symptom: An interval becomes subtraction. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Clarify the range relation or use local range typography. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Are these endpoints or operands? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 29: Decimal punctuation mistaken

Symptom: A grouping mark becomes a decimal separator. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Localise the number as data with correct locale rules. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: What numeric value does it encode? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 30: Abbreviation boundary misread

Symptom: A period after an abbreviation is treated as sentence end. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Use context-sensitive segmentation and review. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Where does the sentence actually end? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 31: Question becomes statement

Symptom: Interrogative punctuation disappears. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Preserve the speech act or supply appropriate question construction. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Is an answer requested? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 32: Exclamation overdramatized

Symptom: Neutral emphasis becomes an emergency warning. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Match pragmatic intensity to original intent. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Does the notice imply new urgency? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 33: Heading and body merged

Symptom: A title fragment becomes the first part of a content sentence. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Keep heading structure and sentence completeness separate. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Where does the heading end? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 34: Citation superscript drift

Symptom: A reference appears to support the next assertion. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Reanchor the citation and test reading order. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: Which claim does the source support? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Failure 35: Copy-paste mismatch

Symptom: Rendered text looks correct but copied logical order is wrong. This is a translation and production defect only when it changes an actual distinction the source or target reader needs. Diagnose the semantic effect first, then inspect whether it originated in translation memory, markup, typographic normalisation, document conversion, or the final rendering engine.

Correction: Test copied Unicode sequence and inspect bidi handling. Do not merely substitute one punctuation mark for another and assume the result is fixed. Verify the underlying subject, action, scope, reference or data value against the source. The necessary output can be a rephrased sentence, a structural list, an attribution label, a corrected Unicode character, or a rendering-specific isolation rule.

Acceptance question: What text is actually exported? The target should answer this accurately without requiring the reader to reverse-engineer the original page. A reviewer can use a before-and-after contrast and record the expected reading. For high-risk material, inspect both displayed text and extracted text; a visual repair that corrupts accessible or copied output is not a complete repair.

Fifty-four worked diagnostics across eighteen everyday communication settings

Every case is fictional and intentionally simplified. The source is an English-language meaning statement, the unsafe form shows an editorial failure, and the repair describes the distinction to preserve in another language. The task is not to reproduce English punctuation internationally: it is to reconstruct the same interpretation using the target language’s grammar, typography and document conventions. The final check should be made with the rendered publication in its actual medium.

Domain 1: Interviews and oral history

A transcript distinguishes voice and interruption. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 1: Interviews and oral history — case 1

Fictional source meaning: “The visitor said, “I saw a light.” The historian then added a caveat.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The historian saw a light.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the visitor’s statement inside its own quotation and place the historian’s caveat outside. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Who personally observed the light? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 2: Interviews and oral history — case 2

Fictional source meaning: “She said, “My uncle told me ‘the river changed course.’”.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “She observed the river change course.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep both the narrator and uncle as separate information sources. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: How many speakers are in the chain? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 3: Interviews and oral history — case 3

Fictional source meaning: “The narrator began “I was going to—” and stopped.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The narrator promised to go.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the interrupted utterance without inventing its completion. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What action was actually promised? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Interviews and oral history: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 2: Research papers

Punctuation supports the relationship between result and interpretation. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 4: Research papers — case 1

Fictional source meaning: “The assay indicated a change; the cause remains undetermined.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The assay established the cause.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Separate measured change from uncertain explanation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which proposition has empirical support? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 5: Research papers — case 2

Fictional source meaning: “The authors wrote “suggestive evidence,” then listed limitations.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The authors proved the effect.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Retain quotation and qualification near the relevant inference. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which conclusion did the authors endorse? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 6: Research papers — case 3

Fictional source meaning: “Results were consistent with two models: A and B.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Only model A was verified.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the colon’s enumerating function and both model alternatives. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does the source exclude the second model? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Research papers: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 3: Technical procedures

A list structure may determine the order and conditions of work. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 7: Technical procedures — case 1

Fictional source meaning: “Step 1: isolate power. Step 2: inspect the enclosure.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Inspect while power remains connected.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep numbered order and the prerequisite relationship. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What action occurs first? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 8: Technical procedures — case 2

Fictional source meaning: “Use either method A or method B; not both.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Use method A and method B.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the exclusive alternative rather than converting it to a cumulative list. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Are both methods permitted together? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 9: Technical procedures — case 3

Fictional source meaning: “Perform the final check—only after the previous tests pass.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Perform the final check immediately.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the dash’s precondition or replace it with a clear target-language clause. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: When may the final check begin? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Technical procedures: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 4: Product interfaces

Compact text creates high risk of implying the wrong action. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 10: Product interfaces — case 1

Fictional source meaning: “Remove account… opens a confirmation dialogue.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Remove account executes immediately.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Use a control label and explanatory context that reflects the actual next step. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does clicking complete the operation? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 11: Product interfaces — case 2

Fictional source meaning: “Change language: English, Arabic, French.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Change language to all three at once.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Represent the options as alternatives in a semantic selection control. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: How many languages may be selected? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 12: Product interfaces — case 3

Fictional source meaning: “The warning says “unsaved changes”; the user must choose Save or Discard.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The warning states changes were saved.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep quoted warning and alternative actions distinct. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Has any saving occurred? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Product interfaces: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 5: News briefs

Headlines and quotations can change who owns a claim. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 13: News briefs — case 1

Fictional source meaning: “An official said, “We are reviewing the report.”.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The report has been confirmed by officials.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep reviewing status and the speech attribution. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Has an official determination occurred? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 14: News briefs — case 2

Fictional source meaning: “A headline quotes “breakthrough” as a contested label.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The article endorses a breakthrough as proven.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the editorial distance conveyed by the quoted term. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Is the claim adopted or questioned? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 15: News briefs — case 3

Fictional source meaning: “The full report says an event was possible—not established.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The event happened.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the dash-supported contrast or equivalent full sentence. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What part remains uncertain? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from News briefs: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 6: Business documents

Parentheses and semicolons can carry operative exceptions. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 16: Business documents — case 1

Fictional source meaning: “Deliver within five days (subject to written approval).” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Delivery is guaranteed within five days.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the approval condition with the delivery promise. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Is performance unconditional? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 17: Business documents — case 2

Fictional source meaning: “Charges apply to A; B is excluded.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Charges apply to both A and B.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep independent clauses and their different treatments. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which item is excluded? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 18: Business documents — case 3

Fictional source meaning: “The proposal includes: design, review and optional testing.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “All three activities are compulsory.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep optional status attached only to testing. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What is required and what is optional? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Business documents: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 7: Museum labels

A label may distinguish historical source from curator interpretation. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 19: Museum labels — case 1

Fictional source meaning: “An archive records “blue vessel”; the object is greenish today.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The original object was definitively blue.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Retain quoted catalogue wording as historical evidence. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Who described the colour? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 20: Museum labels — case 2

Fictional source meaning: “This painting is attributed to a workshop (not an individual artist).” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A named artist made the painting.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Retain the boundary around the attribution. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What remains unresolved? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 21: Museum labels — case 3

Fictional source meaning: “The exhibition offers three readings: trade, ceremony and household use.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Only one use is established.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Present the alternatives without promoting them to facts. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Do the curators select a final interpretation? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Museum labels: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 8: Customer support

Quoted customer speech must not become a confirmed diagnosis. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 22: Customer support — case 1

Fictional source meaning: “The caller reported, “The screen turns blank after login.”.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A system defect after login is confirmed.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the caller’s report distinct from verified diagnosis. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Has the symptom been reproduced? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 23: Customer support — case 2

Fictional source meaning: “A message says “replacement pending”—no replacement shipped.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The replacement has been dispatched.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the status contrast as an incomplete process. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What has actually happened? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 24: Customer support — case 3

Fictional source meaning: “Option 1: restart the application. Option 2: request assistance.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Both actions are mandatory.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep alternate routes and their label scope. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Can the user select one option? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Customer support: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 9: Public alerts

Punctuation must not convert possibility or condition into permission. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 25: Public alerts — case 1

Fictional source meaning: “Do not enter: inspection in progress.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Entry is allowed during inspection.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the prohibition and reason as distinct but linked statements. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Is entry permitted? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 26: Public alerts — case 2

Fictional source meaning: “All-clear not issued—conditions are being reviewed.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “All clear has been issued.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the explicit negation rather than relying on a vague dash. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Who authorised an all-clear? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 27: Public alerts — case 3

Fictional source meaning: “A list says report location, time and observed smoke (if safely possible).” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Every person must approach the smoke to record it.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the safety qualification attached to the observation request. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What should never be inferred? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Public alerts: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 10: Education assessment

Commands, item numbers and examples should remain distinguishable. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 28: Education assessment — case 1

Fictional source meaning: “Question 4(a): explain a cause. Question 4(b): evaluate evidence.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Answer part b by only listing causes.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep question part identities and different command verbs. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which skill is requested in each part? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 29: Education assessment — case 2

Fictional source meaning: “Choose one: explanation, diagram or table.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Provide all three formats.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the one-option instruction. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: How many outputs are requested? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 30: Education assessment — case 3

Fictional source meaning: “The rubric states “for example” before a list.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Only the examples on the list are permitted.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the illustrative status of examples. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Is the list exhaustive? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Education assessment: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 11: Data and reports

Symbols can be data operators or formatted prose. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 31: Data and reports — case 1

Fictional source meaning: “A change of −5 means the value decreased.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A change of 5 means an increase.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the negative sign and its mathematical role. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What is the direction of change? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 32: Data and reports — case 2

Fictional source meaning: “An interval spans 3–5 units.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The data subtract five from three.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Represent endpoints as a range not an arithmetic operation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Is this a range or a calculation? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 33: Data and reports — case 3

Fictional source meaning: “The figure is 1,250.5 in a specified locale.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The figure is read as a different magnitude.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Use a correct target-locale representation while keeping the numeric value unchanged. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What numerical quantity is encoded? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Data and reports: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 12: Multilingual directories

Mixed-direction strings contain neutral punctuation and embedded identifiers. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 34: Multilingual directories — case 1

Fictional source meaning: “An RTL contact name is followed by Latin email and digits.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The email characters appear scrambled or attached to the wrong name.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep logical order and isolate directional runs at the correct boundary. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which person owns the email? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 35: Multilingual directories — case 2

Fictional source meaning: “An Arabic line includes a product code such as AB-15.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The code is reversed to 51-BA.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve stored identifier order and use appropriate direction markup. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What is the actual product code? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 36: Multilingual directories — case 3

Fictional source meaning: “A database cell contains an RTL title and English quotation punctuation.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The closing quote appears beside the wrong term.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Inspect embedded-run isolation rather than manually reversing characters. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which phrase is quoted? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Multilingual directories: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 13: Poetry and plays

Hard breaks and speaker turns may carry interpretation. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 37: Poetry and plays — case 1

Fictional source meaning: “Line one ends with “not”; line two begins with the named action.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The poem is reformatted as a straightforward positive statement.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve the meaningful enjambment or provide a carefully justified target effect. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What tension does the break create? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 38: Poetry and plays — case 2

Fictional source meaning: “The script uses a dash for a new speaker.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The continuation is attributed to the previous actor.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve speaker turn labels across translation and layout. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which character speaks? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 39: Poetry and plays — case 3

Fictional source meaning: “A character trails off with an ellipsis before answering.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The actor gives a crisp immediate yes.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Preserve hesitation if it is meaningful to performance. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What does the pause communicate? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Poetry and plays: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 14: Online collaboration

Threads and nested comments can lose reply scope. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 40: Online collaboration — case 1

Fictional source meaning: “A reviewer writes “approved” only beneath the revised heading.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The reviewer approved the whole document.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep comment attachment and scope visible. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What exactly was approved? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 41: Online collaboration — case 2

Fictional source meaning: “A code review contains a quoted error message and an editor’s diagnosis.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The system itself issued the editor’s diagnosis.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep log text distinct from human interpretation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which words came from the machine? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 42: Online collaboration — case 3

Fictional source meaning: “A task card lists blocked items as sub-bullets of a parent issue.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “All tasks are independently unblocked.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the nested structure and blocker relation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which parent remains blocked? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Online collaboration: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 15: Healthcare instructions

The target must not invent permissions or probabilities. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 43: Healthcare instructions — case 1

Fictional source meaning: “The sheet says “take only if instructed by your clinician”.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “Take routinely without consultation.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the conditional language and do not interpret it as personal advice. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Who authorises the action? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 44: Healthcare instructions — case 2

Fictional source meaning: “The note says “patient-reported” discomfort, then records observations.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “All statements are measured findings.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep source framing distinct from clinician findings. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Who supplied each statement? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 45: Healthcare instructions — case 3

Fictional source meaning: “An exception in parentheses applies to one measurement group.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The exception applies to every patient.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Restore the intended scope and obtain clinical review. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which subgroup is affected? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Healthcare instructions: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 16: Legal records

Quotation and clause punctuation affect attribution and operative scope. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 46: Legal records — case 1

Fictional source meaning: “The pleading alleges breach; the ruling reserves judgment.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The court found a breach.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep allegation and decision as separate acts. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What was legally determined? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 47: Legal records — case 2

Fictional source meaning: “A contract defines “service day” in a colon-led definition.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The translator uses the ordinary calendar meaning instead.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the defined term and all limits specified by its clause. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What does the document define? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 48: Legal records — case 3

Fictional source meaning: “A footnote qualifies only section 3(a).” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The footnote changes all contract sections.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep reference anchor and restricted section scope. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Which paragraph is qualified? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Legal records: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 17: Accessible publishing

Presentation and reading order can diverge. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 49: Accessible publishing — case 1

Fictional source meaning: “A visually indented list is stored as plain paragraphs.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A screen reader cannot identify nested conditions.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Use semantic list markup preserving hierarchy. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does assistive technology announce the structure? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 50: Accessible publishing — case 2

Fictional source meaning: “A line-wrapped link is split into two unrelated fragments.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A keyboard user cannot activate the complete link.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep the link as one logical element and allow safe wrapping. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: What is the actual clickable target? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 51: Accessible publishing — case 3

Fictional source meaning: “The site copies a multilingual title with invisible direction controls.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The copied title no longer matches the visible title.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Inspect logical string direction controls and extraction. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does copy and search preserve identity? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Accessible publishing: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Domain 18: Automated localisation

Software processing can erase or invent boundaries. Translators should begin by writing a short structural outline of the content rather than copying graphic marks from the source. Identify the actor, action, condition, quotation owner and any reference or data token that must remain stable. The right translation is not always the one with identical punctuation; it is the version from which an ordinary target reader can infer the same relationships.

Diagnostic 52: Automated localisation — case 1

Fictional source meaning: “A translation pipeline strips quote marks while preserving words.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “An attributed sentence becomes an editorial assertion.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Treat quotation boundaries as QA invariants. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Can the speaker still be identified? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 53: Automated localisation — case 2

Fictional source meaning: “A localisation engine replaces a mathematical minus with a hyphen.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “A parser reads the negative field incorrectly.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Protect data operators from prose punctuation normalisation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does the parsed number remain equal? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Diagnostic 54: Automated localisation — case 3

Fictional source meaning: “A formatter adds hard newlines at screen wrap positions.” Read this as a complete communication act. Classify the role of each punctuation mark or layout boundary: does it introduce a quotation, govern a clause, separate choices, delimit a data token or indicate a point in a process? Distinguish marks that a house style may change from marks whose function must survive translation.

Unsafe localisation: “The exported document gains extra breaks.” The flaw is not just that the line is less elegant than the source. It changes the information available to readers, often by moving a condition, inventing an endorsement or treating display order as the actual text sequence. A translation-quality reviewer must name the specific mistaken interpretation and show which source boundary prevented it.

Meaning-preserving repair: Keep soft wrap out of the content representation. In the target language, this may require different quotation marks, a clearer clause, a proper list, a revised emphasis pattern or programmatic direction isolation. Avoid adding facts or stronger requirements merely to make the sentence concise. If the source itself has two plausible readings, submit a focused author query rather than choosing one invisibly.

Reader test: Does a narrower display alter stored content? Show the rendered target to someone who has not read the source. Ask them to answer this question and describe which typography gave them that impression. If the test answer differs from the source’s intended interpretation, correct the target and rerun the test on the shortest format in which the sentence will appear. Short excerpts and notification cards often lose essential qualifiers first.

Practical transfer from Automated localisation: take the same three source distinctions and encode them as a normal paragraph, a narrow-screen summary and a structured record. All three formats need the same boundaries even when their punctuation looks different. Compare copy-and-paste output, accessibility order and printed rendering when the material will circulate through multiple technologies. Log any intentional variation in style so that future updates do not mistake a valid difference for an error.

Independent transfer workshops

Transfer workshop 1: Dialogue quotation

Task: Translate an interview with two nested quotation levels. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Keep four ownership boundaries even if the target convention uses different delimiters. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 2: Contract exception

Task: Put a single-item exclusion into a multi-level clause. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: A reader must identify which item carries the exception. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 3: Scientific estimate

Task: Move a negative value and a range between prose and a table. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Keep the mathematical quantity independent of typography. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 4: RTL product title

Task: Insert an opposite-direction product name with a Latin code into an RTL sentence. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Preserve logical order and isolate the embedded run. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 5: Subtitles

Task: Condense a hesitant interrupted utterance into a two-line subtitle. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Preserve an unfinished promise rather than creating a completed commitment. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 6: Short alert

Task: Reformat a prohibition plus an explanation into a mobile notification. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The prohibition must not be hidden in the second line. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 7: Archive quotation

Task: Quote an old report with an editorial omission. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Mark omitted words without inventing continuity. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 8: Survey choice

Task: Move one-of-three alternatives from a paragraph into a form. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Keep alternatives separate and do not turn them into cumulative requirements. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 9: Legal definition

Task: Reproduce a defined term introduced by a colon. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The target definition must apply to the same concept throughout the document. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 10: Poetic break

Task: Translate a two-line contrast with a deliberate pause. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Preserve the meaningful opposition rather than accidental visual wrapping. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 11: Service ticket

Task: Differentiate machine output from a quoted customer statement. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The reader must know who generated each string. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 12: Accessibility

Task: Create semantic heading and list structure from visually styled text. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Screen-reader order must match visual groupings. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 13: Page title

Task: Localise quoted book titles next to quoted speech. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The title must not become a speaker’s remark. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 14: Address

Task: Break a multi-line postal address safely for narrow screens. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The full address must remain recoverable in logical order. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 15: Financial table

Task: Translate grouping and decimal separators for two target locales. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The numerical value must be identical after parsing. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 16: Historical caption

Task: Distinguish a curator’s gloss from a quoted historical label. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The editor must not appear to have authored the older words. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 17: Footnote

Task: Attach one explanatory note to a specific claim in a long paragraph. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The anchor must not drift during translation. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 18: Procedure

Task: Keep a numbered sequence when one step is optional. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Optional status must stay attached to its step. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 19: Message repair

Task: Represent an abrupt dash interruption without a false apology or agreement. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The speech act must remain incomplete. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 20: Character truncation

Task: Shorten a line containing combining marks and emoji sequences. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Never cut inside a perceived character cluster. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 21: Citation

Task: Move a citation through a multilingual table and paragraph. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: The source must remain attached to the claim it supports. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 22: Rich-text export

Task: Export a mixed-script quotation to plain text and PDF. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Logical order and quotation scope must survive both formats. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 23: Search snippets

Task: Produce a short page description containing a cautious claim. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: A search reader must not infer stronger certainty than the source. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 24: Update diff

Task: An editor removes a comma while leaving every word unchanged. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Review whether syntactic scope has shifted. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Transfer workshop 25: Glossary

Task: Create a style guide entry for apostrophes and hyphens. Begin with a sentence-level meaning map and identify which text boundaries are semantic, conventional and rendering-specific. Choose the target-language construction first and the exact graphic characters second; this avoids the common mistake of translating a symbol before understanding what the symbol does.

Pass criterion: Record meaning exceptions rather than one blanket substitution. A second reviewer should try to paraphrase the output without seeing the original. If that interpretation diverges, mark whether the issue arose through grammar, typography, layout, directionality or extraction and revise that layer. Keep both the repaired target and an explanation of why it preserves the meaning, so the test becomes reusable during future releases.

Quality gates for live publication

The most useful quality audit is a two-layer comparison. First, verify linguistic meaning: quotation, clause scope, actor, action, sequence, emphasis and numerical interpretation. Second, verify the technical representation: Unicode characters, grammatical segmentation, rendered line breaks, stored order, copy-and-paste output and assistive-technology reading route. A document can be correct at one layer and fail at the other. Two independent checks avoid treating a beautiful screenshot as proof that the underlying text is correct.

Develop a risk register for the document’s highest-consequence patterns. Quotations from identifiable individuals, negative numbers, exception lists, binding conditions, addresses, product codes, citations and warnings all deserve explicit positive tests. A minor variation in decorative spacing can be accepted when a style guide permits it; a change to who is quoted or which condition applies cannot be dismissed as mere house style. Define the approval authority and record whether the correction is linguistic, technical or both.

Review the output in the ways real people will consume it. A web page needs browser and viewport checks; a PDF needs reading-order and extraction checks; a subtitle needs timing and speaker alignment; a form needs keyboard navigation and labels; a long document needs navigable heading, list and table structure. Do not infer these are solved simply because the same source strings appear in all formats. Export pipelines and renderers can introduce new failures after the translation itself has been approved.

A practical release record contains the source identifier and revision, target language, target edition, project style guide, list of known high-risk spans, test devices or outputs, error decisions, correction owner and approval date. When a later revision changes a comma, a quote or a single newline, the affected target segment may require renewed review even if the words themselves look unchanged. The objective is a recoverable interpretation, not a superficially identical page.

A four-week practice plan

Week 1: recover meaning boundaries

Collect short examples of direct quotation, nested speech, comma scope, list hierarchies, number tokens and interrupted dialogue. For each, write a plain-language explanation of what the punctuation does. Ask a second reader to challenge your explanation using a deliberately modified version. The outcome should be a small ledger of boundaries that must survive any translation.

Keep one before-and-after example for each repaired error. A usable training result records exactly which reader inference was wrong, why the mark or layout caused it, and how the final text prevents it. When that record is understandable to another translator or engineer, the skill has transferred beyond the original exercise.

Week 2: localise convention without moving scope

With a competent reviewer for the actual target language, translate the examples using its punctuation and stylistic conventions. Where a symbol does not transfer naturally, replace it through syntax or structural markup. Re-read the target alone and compare the quotation ownership, options, conditions and data values with the original. Do not assume a one-to-one punctuation map.

Week 3: test rendering and extraction

Move approved text into a web page, narrow mobile view, exported PDF and plain-text copy. Include contrasting left-to-right and right-to-left runs, digits, parentheses and citations. Test grapheme-safe selection, line wrapping and accessible heading or list structure. Classify each newly discovered failure by layer, then rerun a small regression suite after the correction.

Week 4: transfer and maintain

Give an independent reviewer a fresh document containing unfamiliar punctuation and layout structures. Require a rationale for ambiguous source boundaries, a corrected target, a rendered-output check and a version record. Simulate a small source edit that changes one quote boundary or list item and verify that the updated target preserves the new meaning without breaking other references.

Glossary of practical punctuation and text-engineering terms

Logical text order

The character sequence stored in the source or data model before bidirectional display processing. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Visual order

The arrangement of glyphs after text direction layout and font shaping. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Bidirectional text

Text containing runs whose normal reading directions differ. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Base direction

The paragraph or isolated run direction used to interpret mixed text. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Direction isolation

A technique preventing an embedded directional run from unexpectedly affecting surrounding text. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Unicode code point

An abstract character identity addressed by a Unicode scalar value where applicable. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Glyph

A visible form chosen by a font for rendering text. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Grapheme cluster

A sequence treated as a user-perceived written character under a segmentation profile. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Combining mark

A character that contributes an accent or other feature to a base sequence. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Normalisation

A defined transformation between canonically or compatibly equivalent Unicode representations. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Line-break opportunity

A position where wrapping may be permitted under a line-breaking rule. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Mandatory break

A location that requires separation under a relevant line-breaking rule. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Soft wrap

A display-created line ending that is not a semantic newline. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Hard line break

A stored break deliberately separating lines within content. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Paragraph boundary

A structural separation between paragraph-level text units. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Nonbreaking space

A space character that prevents a break at that position. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Quotation boundary

The start or end of text represented as another speaker’s or document’s words. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Nested quotation

Quotation embedded within another attributed quotation. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Scare quotes

Quotation marks signalling distancing special usage or contestability rather than verbatim speech. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Direct speech

Words attributed as being spoken or written by a particular person or source. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Reported speech

Content conveyed through a narrator or paraphrase instead of represented verbatim. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Ellipsis

A sequence marking omission hesitation continuation or another context-specific function. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Editorial interpolation

Text inserted by an editor for clarification and distinguished from original quoted wording. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Colon

A punctuation mark that can introduce an explanation enumeration definition or quotation. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Semicolon

A separator often joining related independent units or complex list members. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Dash

A punctuation or dialogue mark whose kind and function depend on context. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Hyphen

A mark typically used in orthography word formation or line division. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Minus sign

A mathematical operator distinct from an orthographic hyphen. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Range delimiter

A mark or phrase expressing endpoints rather than subtraction. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Restrictive modifier

A phrase selecting which member of a broader group is intended. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Nonrestrictive modifier

A phrase adding supplementary information about an already identified referent. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

List hierarchy

The parent child structure connecting list entries and subentries. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

List item scope

The content governed by a particular bullet or numbered unit. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Footnote anchor

A reference linking a source span to its explanatory note. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Citation scope

The proposition or passage supported by a cited source. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Semantic markup

Structure encoding document roles independently of graphic appearance. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Accessible reading order

The sequence through which assistive technology presents the content. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Text extraction

Recovery of stored words and structure from a displayed or exported medium. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Copy-and-paste integrity

Preservation of logical character sequence when text moves between applications. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Typographic convention

A language and medium specific practice for punctuation spacing or layout. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

House style

An organisation’s adopted editorial conventions for consistency. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Rendering engine

Software that converts text and markup into displayed glyphs and layout. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Bidirectional isolate controls

Unicode characters designed to isolate directional runs where markup is unavailable. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Regression test

A repeated test guarding against reintroduction of a previously fixed error. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Semantic invariant

A meaning relationship required to survive language and format changes. For translation review, record whether this is a semantic relation, an editorial convention, a data property or an implementation concern. The categories sometimes interact, but they are not synonyms. Testing the correct layer prevents teams from spending time adjusting punctuation while missing the underlying loss of meaning.

Frequently asked questions

1. Should translated punctuation match the source exactly?

No. Preserve what the punctuation communicates, then apply target-language conventions. Exact graphic correspondence may be wrong when languages use different systems.

Test the answer on a short source passage rather than memorising a fixed character substitution. Mark what the source means, choose the target’s natural representation and check the interpretation of someone reading only the result. A single successful visual comparison is not a substitute for the logical-text or reading-order check when those layers are relevant.

2. When do quotation marks change meaning?

They can determine who owns a claim, distinguish titles from dialogue, or signal that a term is contested. Review quotation scope before normalising marks.

3. Can I replace a semicolon with a full stop?

Often yes, if the relationship between the clauses remains accurate and no condition, list grouping or interpretation changes.

4. How should an ellipsis in a translated quotation be handled?

Determine whether it marks editorial deletion, a speaker’s hesitation or another function. Use the target publication’s conventions without concealing omitted source material.

5. What is the difference between a hyphen and a minus sign?

A hyphen commonly has orthographic roles; the minus sign is a mathematical operator. Confusing them can change how a numeric expression is read or parsed.

6. Does an en dash always mean a range?

No. Its typographic role depends on the language and style guide. Check the actual relation instead of applying a blanket rule.

7. How can commas affect legal or technical meaning?

Commas can influence clause attachment, restriction and exception scope. A wrong boundary may change who or what a rule covers.

8. Why can line breaks matter in translation?

Intentional breaks can separate dialogue, verse, address components or procedural units. Automatic wrapping is different and should not become stored content.

9. What is the difference between a hard break and soft wrap?

A hard break is stored structure; a soft wrap is a presentation decision caused by available width or layout.

10. Why does mixed Arabic and English punctuation sometimes look displaced?

The Unicode bidirectional algorithm lays out mixed directional runs. Neutral punctuation may attach visually to an unexpected run if direction boundaries are not marked appropriately.

11. Should I reverse a right-to-left text string manually?

No. Preserve logical character order and use appropriate markup or bidirectional controls. Manual reversal can corrupt copying, search and assistive reading.

12. When should I use direction isolation?

When an embedded opposite-direction phrase might affect punctuation or surrounding runs. W3C inline bidi guidance explains robust HTML patterns.

13. What is a grapheme cluster?

A user-perceived written character that may include several code points. Text editing and truncation often need to respect these boundaries.

14. Are code points the same as visible characters?

Not always. Combining marks, sequences and font shaping can produce one perceived character from multiple code points.

15. Does Unicode dictate every language’s punctuation style?

No. Unicode defines characters and algorithms, while editorial conventions vary by language and publisher.

16. Does the line-breaking algorithm choose each displayed line?

It supplies break opportunities and related rules. Layout width, font metrics and higher-level software determine many actual breaks.

17. What if a list mixes instructions and examples?

Preserve which entries are required, optional, illustrative or alternative. If the source is ambiguous, ask its author rather than creating obligations.

18. What is the safest way to translate a nested list?

Represent the hierarchy explicitly, map the scope of each child item and test the rendered reading order.

19. Can visually identical punctuation have different Unicode identities?

Yes. The renderer may make distinct characters look similar. Inspect code points where exact data identity matters.

20. What should be checked in PDF export?

Quotation boundaries, list hierarchy, text extraction, logical reading order, fonts and the placement of footnotes or reference marks.

21. What should be checked in subtitles?

Speaker turns, timing, intentional hesitation and meaningful line breaks, without claiming that visual wraps are new speech.

22. Can automated translation change punctuation incorrectly?

Yes. Models and localisation pipelines may alter quotes, list breaks, separators and directional markup. High-risk structures require tests.

23. Should an untranslated identifier be treated as ordinary text?

Usually not. Protect its identity, numeric signs, ordering and code points while localising the surrounding explanation.

24. How do translators handle punctuation absent from the target script?

Reconstruct the intended meaning through the target’s natural syntax, punctuation or layout; avoid foreign punctuation imposed solely for resemblance.

25. How can a reviewer identify punctuation-related risk quickly?

Focus on quotes, exception lists, conditions, data tokens, mixed-direction names, headers and citations. Ask what a wrong boundary would cause a reader to do.

26. Is a screenshot proof of correct multilingual rendering?

No. The logical sequence, copy result, accessibility route and behaviour on other screen widths still need checking.

27. How can we preserve citations in translated text?

Keep note anchors and reference identifiers attached to the claims they support; validate links and exports after layout changes.

28. What if an old translation is already published with wrong punctuation?

Correct the affected text, document the substantive distinction and retest its exported and responsive versions. Do not assume a visual-only patch is enough.

29. Where should beginners learn routine punctuation translation?

Use the existing introduction to punctuation, quotations and capitalisation. This chapter addresses deeper semantic and technical integrity across publishing systems.

30. What is the final publication test?

Show a source-free target to a reader and ask who says what, what is omitted, which actions are conditional and what numeric or identifier strings are present.

Verified reference sources and standards

Unicode Standard Annex #9: Unicode Bidirectional Algorithm — Unicode 18.0.0, revised 1 September 2026. This provides the text-display ordering model for mixed-direction characters, not translation semantics.

Unicode Standard Annex #14: Unicode Line Breaking Algorithm — Unicode 18.0.0, revised 1 September 2026. This describes line-break opportunities and constraints, while actual layout depends on higher-level software.

Unicode Standard Annex #29: Unicode Text Segmentation — Unicode 18.0.0, revised 1 September 2026. It defines default grapheme, word and sentence boundary procedures that may be tailored for language-specific needs.

W3C Internationalization: Inline Markup and Bidirectional Text in HTML — practical direction-markup and isolation guidance for embedded opposite-direction runs, numbers, punctuation and names.

W3C Internationalization: Approaches to Line Breaking — explains differences in wrapping conventions among scripts and why punctuation-sensitive breaks cannot be assumed identical.

Unicode Standard Annex #15: Unicode Normalization Forms — useful when character representation and text processing influence matching, without implying that all typographic distinctions are disposable.

Unicode Common Locale Data Repository — a source of locale data; apply relevant locale rules alongside a suitable editorial style guide rather than assuming English punctuation is universal.

Next routes in the translation architecture

Architecture reference: EDKSG-TRANS-MASTER-WORLD-1370. Return to Master Art of Translation | The Complete System for Moving Meaning Between Languages for the complete translation system. The previous child, The Entity Identity Preservation System, explains how names, scripts, aliases and identifiers retain identity; this chapter continues from stable character identity into punctuation, structure and rendered meaning.

For the existing introductory explanation, use How to Translate Easily | How to Translate Punctuation, Quotation Marks and Capitalisation Naturally. For the dedicated standards-focused treatment of language and direction metadata, use The W3C String Language and Direction Metadata System. These articles remain distinct owners and are not replaced by this chapter.

For underlying word concepts and language learning, use the protected Vocabulary Learning Hub. For English sentence interpretation, quotation and grammar, use How English Works | Translation and Meaning Across Languages. These are complementary routes, not content to be duplicated or overwritten.

The final measure is whether a reader can recover the same actors, claims, conditions, omissions, quantities and reading route from the target as from the source. Typography should help that work, not become an invisible second author. When both language and representation have been checked, the result can change its script and layout without changing what it actually says.

Master Art of Translation | The Typography and Punctuation Meaning System — How Quotation, Ellipses, Lists, Line Breaks and Text Direction Keep Meaning